Overview on MIMIC-III

[JohannesWiesner]

I am interested in working with EHR data from psychiatric patients (specifically patients with schizophrenia). The MIMIC-III database seems very promising to me, however, it is not clear to me how to get a rough overview of this dataset and if it fits my needs. Is there an 'overview' feature that does this for me? I miss something like a frequency chart for psychiatric ICD-10 diagnoses to give me intuition about the number of psychiatric patients.

[tompollard]

There is no publicly available frequency chart for psychiatric ICD (-9) diagnoses in MIMIC, as far as I'm aware. You could explore the MIMIC-III demo to see whether you can find the information that you're interested in (https://physionet.org/content/mimiciii-demo/1.4/), but the small patient number might limit your insights. If you list some ICD-9 codes here, someone reading the issue may find the time to take a look.

[JohannesWiesner]

The demo dataset contains only data for "all intensive care unit (ICU) stays for 100 patients". So it is not useful in my case (i.e. get an overview about the number of psychiatric patients and their ICD-9 Codes: 290-319). For me, it would be very helpful to get an overview of ICD-diagnoses so that I can decide in advance if it makes sense for me to take the registration course.

For example, the signature databank offers such an overview.

[tompollard]

As a start, here is a quick query to pull out patients with this range of ICD-9 codes (ordered by number of distinct patients):

SELECT a.icd9_code, b.long_title, count(distinct a.subject_id) as n_patient, 
FROM `physionet-data.mimiciii_clinical.diagnoses_icd` a
LEFT JOIN `physionet-data.mimiciii_clinical.d_icd_diagnoses` b
ON a.icd9_code = b.icd9_code
WHERE a.ICD9_CODE BETWEEN '290%' and '319%'
GROUP BY a.icd9_code, b.long_title
ORDER BY count(distinct a.subject_id) DESC
LIMIT 40;

icd9_code	long_title	n_patient
3051	Tobacco use disorder	3001
311	Depressive disorder, not elsewhere classified	2926
30000	Anxiety state, unspecified	1452
2930	Delirium due to conditions classified elsewhere	1398
2948	Other persistent mental disorders due to conditions classified elsewhere	1026
30500	Alcohol abuse, unspecified	954
3004	Dysthymic disorder	928
29181	Alcohol withdrawal	675
30391	Other and unspecified alcohol dependence, continuous	653
30390	Other and unspecified alcohol dependence, unspecified	568
29410	Dementia in conditions classified elsewhere without behavioral disturbance	551
29680	Bipolar disorder, unspecified	507
30501	Alcohol abuse, continuous	388
30560	Cocaine abuse, unspecified	368
29281	Drug-induced delirium	361
30301	Acute alcoholic intoxication in alcoholism, continuous	301
2910	Alcohol withdrawal delirium	292
30393	Other and unspecified alcohol dependence, in remission	281
30401	Opioid type dependence, continuous	251
30981	Posttraumatic stress disorder	229
2967	Bipolar I disorder, most recent episode (or current) unspecified	187
29590	Unspecified schizophrenia, unspecified	182
30590	Other, mixed, or unspecified drug abuse, unspecified	177
30550	Opioid abuse, unspecified	174
2920	Drug withdrawal	167
29570	Schizoaffective disorder, unspecified	163
29620	Major depressive affective disorder, single episode, unspecified	158
29420	Dementia, unspecified, without behavioral disturbance	158
3079	Other and unspecified special symptoms or syndromes, not elsewhere classified	153
29040	Vascular dementia, uncomplicated	136
2939	Unspecified transient mental disorder in conditions classified elsewhere	130
319	Unspecified intellectual disabilities	129
31401	Attention deficit disorder with hyperactivity	121
30503	Alcohol abuse, in remission	112
2989	Unspecified psychosis	107
30001	Panic disorder without agoraphobia	104
30400	Opioid type dependence, unspecified	102
30520	Cannabis abuse, unspecified	95
29650	Bipolar I disorder, most recent episode (or current) depressed, unspecified	77
30300	Acute alcoholic intoxication in alcoholism, unspecified	77

[JohannesWiesner]

Thanks, that's a good start and in exactly what I wanted. Of course, it would still be practical, if this overview could be implemented as a feature in the future.